| Having an accurate and complete map showing road information is the basis for mobile robots to complete navigation tasks.Among the current mainstream methods,the method of using SLAM technology to build a metric map has high requirements on the storage and computing capabilities of local equipment,while traditional topological maps lack the ability to express the actual environment of the road network.Therefore,it is not necessary to use any map alone.It is difficult to meet the needs of equipment miniaturization and intelligence.In response to the above problems,this paper adopts the technology of 3D point cloud to perceive the road environment,and focuses on the two application scenarios of mapping and navigation,and the following three aspects of work are carried out:(1)Aiming at the outdoor road environment,an intersection detection algorithm based on a single frame of sparse point cloud is designed.The algorithm first removes the interference of the ground point cloud and obstacle point cloud to the detection,and then rasterizes the point cloud to improve the real-time performance of the algorithm.Aiming at the problem of missed detection and false detection of single-frame point cloud detection,a continuous beam model is designed to fully perceive the intersection structure.Experiments show that the algorithm can stably detect intersection information in a road environment with obstacles,and improve the ability to adapt to the environment.(2)Aiming at the limited local computing resources of mobile robots,a waypoint map building framework with intersections as nodes in a road environment is designed.Aiming at the problem of large rotation and translation differences in point clouds obtained from different perspectives at the same intersection,the framework designs a fast and accurate matching algorithm.Aiming at the problem of the instability of single-frame point cloud detection,ajunction discovery algorithm for continuous multi-frame point clouds is designed,and a corresponding pruning strategy is designed to accelerate node recognition during closed loop.Experiments show that the framework can correctly construct a waypoint map in an outdoor road environment and fully express the characteristics of intersections.(3)A mobile robot navigation framework is designed based on the waypoint map.The framework first completes the global planning of the waypoint map,and then designs a local planning algorithm based on intersection detection for the problem that Morphine algorithm has less analysis of road environment information.The local planning algorithm designed in this paper can make full use of the road structure information contained in intersection nodes to assist planning path generation.Navigation experiments in the real road network show that the algorithm can plan a smoother path based on road structure information. |